Watch the build.
Join the conversation.

HN848: NetClaw: Your Open Source Network Assistant

John Capobianco joins Ethan Banks and Drew Conry-Murray for a full NetClaw conversation: MCP, approvals, layered protection, NCFED, specification-driven development, and the people contributing to the project.

5:48 · Soul file and explicit approval ↗
19:48 · NCFED Internet-Draft ↗
23:51 · Layered protection ↗
32:39 · Specification-driven development ↗
44:39 · Eight lines of JSON ↗
1:00:52 · Maintainers and the community ↗

Agentic ServiceNow Remediation: From Alert to Closed Ticket with FlowAgents

John Capobianco and Joksan Flores demonstrate incident coordination, read-only triage, human-approved remediation, and documentation back in ServiceNow.

2:33 · Ticket Churn & Skipped Documentation ↗
8:14 · Recommendations vs. Resolutions & Human Approval ↗
9:44 · The Manual Incident Workflow, Step by Step ↗
13:43 · Coordinator, Triage & Remediation Agents ↗
17:45 · Change Windows & Mixing LLMs per Agent ↗
19:49 · Demo Topology & the FlowAI Agent Harness ↗
23:22 · RBAC Controls on Write-Capable Agents ↗
25:11 · Inside the Agentic Incident Remediation Project ↗
29:42 · Live Demo: Opening a ServiceNow Incident ↗
31:42 · Triage Agent: Read-Only Diagnosis & Audit Trail ↗
37:46 · Work Center Approval & Rollback Plan ↗
42:36 · Remediation Agent: Verify, Apply, Re-Test ↗
45:16 · Incident Summary & ServiceNow Ticket Closure ↗
47:52 · Why Read-Only Triage Is the Place to Start ↗

TCG086: Why Training AI to Have a Self Makes Governance Harder

John Capobianco joins William Collins on The Cloud Gambit to discuss MCP, NetClaw, operational guardrails, specification-driven development, model identity, and specialized decision models.

2:31 · MCP Dev Summit and NetClaw ↗
17:54 · NetClaw’s constitution ↗
25:53 · Specifications and human clarification ↗
47:04 · Model identity and governance ↗
1:00:04 · Specialized decision models ↗

AI, Systems Thinking, and the Future of Network Engineers

Andy Lapteff and guest Ryan Booth discuss AI and systems thinking. Andy recalls John Capobianco’s Hello World for AI session at 1:06:15 and their AutoCon subnetting game at 1:11:02.

1:06:15 · Hello World for AI ↗
1:11:02 · AutoCon subnetting game ↗

Build Network Health FlowAgents Using pyATS MCP & Itential Builder Skills

John connects pyATS MCP to FlowAI and uses Itential Builder Skills to create and test interface-health and device-health agents.

Choose Your Model, Not Your Vendor: Running FlowAgents on Any LLM, Anywhere

John and Joksan show how the Itential Model Registry lets a troubleshooting agent use different hosted and local models while retaining its tools and instructions.

0:54 · Bring Your Own Model & Tokenomics ↗
5:05 · ReAct Agents & Tool Calling Explained ↗
7:20 · FlowAgent Builder: Tool Scoping & RBAC ↗
8:27 · Reasoning + Deterministic Execution: The Spectrum ↗
11:27 · Platform Architecture Refresher ↗
12:23 · Demo: Model Registry Tour ↗
14:46 · On-Prem GPU Connectivity via Gateway (mTLS) ↗
19:53 · Building the Agent: Prompt, Tools & Model Selection ↗
21:25 · Baseline Run: GPT vs. Open Weight Comparison ↗
22:33 · Switching Models: One Edit, Nothing Else Changes ↗
25:02 · Nemotron Ultra: Tool Calling & Agent Visibility ↗
27:06 · Results, Cost Breakdown & Multi-Agent Patterns ↗
31:54 · Regulated Sectors & Wrap-Up ↗